Challenges in predicting PROTAC-mediated protein-protein interfaces with AlphaFold reveal a general limitation on...

Area of Science:

• Structural biology
• Computational chemistry
• Drug discovery

Background:

• Proteolysis Targeting Chimeras (PROTACs) are heterobifunctional molecules inducing targeted protein degradation, offering a promising therapeutic strategy for cancer.
• Accurate prediction of PROTAC-mediated protein complexes is crucial for *in silico* drug design pipelines.
• Previous observations indicated limitations of AlphaFold2 (AF2) in predicting PROTAC-mediated complexes.

Purpose of the Study:

• To investigate the underlying reasons for AlphaFold2's failure in predicting PROTAC-mediated protein complexes.
• To evaluate the impact of interface size and ligand presence on AF2's prediction accuracy.
• To benchmark AlphaFold3 (AF3) against AF2 for PROTAC complex prediction.

Main Methods:

• Analysis of 326 protein heterodimers to assess AF2 performance based on interface size and ligand presence.
• Evaluation of AF2-multimer predictions on complexes with varying interface sizes.
• Benchmarking AF2 and AF3 on a dataset of 28 PROTAC-mediated dimers.

Main Results:

• AF2 predictions are sensitive to interface size, with significant inaccuracies observed for smaller interfaces, even without ligands.
• AF3 did not demonstrate a substantial improvement in accuracy over AF2 for PROTAC-mediated dimer prediction.
• The limitations of AF2 in predicting small interfaces have critical implications for computational PROTAC design and other prediction tasks involving small interfaces.

Conclusions:

• AlphaFold2's accuracy in predicting protein complexes is significantly affected by interface size, particularly for small interfaces.
• The observed limitations pose challenges for *in silico* design of Proteolysis Targeting Chimeras, which often involve small interaction interfaces.
• Further development is needed to improve the accuracy of protein structure prediction tools for complexes with small interfaces, essential for advancing drug discovery pipelines.